Стенд-проверка (stage-2) · промт оператора:

(универсальный)

Ответ агента:

{ "worked": true, "confidence": 8, "verdict": "Built a custom sphere_one_step_dynamics track because the built-in dynamics track does not contain manifold-valued states or exponential-map updates. Using the same mlp_tiny backbone and equal training budget, the exponential-map residual achieved lower test MSE than the swept additive baseline on all 8 paired seeds; the paired permutation test was significant. The trained-model norm signature also quantitatively confirmed the predicted constraint preservation.", "metrics": { "baseline": "Tuned additive Euler, full 8-seed test MSE mean 1.9605001710942815e-06 ± 4.787010942122163e-07; best cfg lr=0.01, h=0.12.", "idea": "Tangent-projected spherical exponential map, full 8-seed test MSE mean 1.8447870697713142e-07 ± 3.449141553482167e-08; best tested setting among 3 idea settings.", "paired_delta": "-1.7760214641171501e-06; idea wins 8/8 seeds; permutation p=0.0081." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_sphere_bench.py", "files": [ "sphere_track.py", "stage2_sphere_bench.py", "bench_report.json", "bench_run.log" ], "limitations": "This is a small custom one-step S2 regression task rather than a large neural world model or stochastic diffusion rollout. Noise learning, metric-induced anisotropic diffusion, transition log-density/Jacobian terms, multi-step horizon behavior, wall-clock/FLOP efficiency, and other manifolds were not tested. The reported significant gain may partly reflect the target's exact unit-sphere geometry and should not be generalized without broader benchmarks.", "bench_report": { "bench_version": 1, "track": "sphere_one_step_dynamics", "model": "mlp_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "h": 0.12 }, "full": { "mean": 1.9605001710942815e-06, "std": 4.787010942122163e-07, "per_seed": [ 1.5833142015253543e-06, 1.5794613545949687e-06, 2.8728609322570264e-06, 1.9276044440630358e-06, 2.232864062534645e-06, 2.1574342099484056e-06, 2.126187155226944e-06, 1.2042750086038723e-06 ], "n": 8 }, "sweep": [ { "cfg": { "lr": 0.001, "h": 0.06 }, "mean": 0.00018889913917519152 }, { "cfg": { "lr": 0.001, "h": 0.12 }, "mean": 8.064255553108524e-05 }, { "cfg": { "lr": 0.001, "h": 0.24 }, "mean": 3.386509808933624e-05 }, { "cfg": { "lr": 0.003, "h": 0.06 }, "mean": 1.4137278640191653e-05 }, { "cfg": { "lr": 0.003, "h": 0.12 }, "mean": 5.718101306229073e-06 }, { "cfg": { "lr": 0.003, "h": 0.24 }, "mean": 3.6321278003015323e-06 }, { "cfg": { "lr": 0.01, "h": 0.06 }, "mean": 2.5334072688565357e-06 }, { "cfg": { "lr": 0.01, "h": 0.12 }, "mean": 1.9908102331100963e-06 }, { "cfg": { "lr": 0.01, "h": 0.24 }, "mean": 2.2787774582866405e-06 } ] }, "idea": { "mean": 1.8447870697713142e-07, "std": 3.449141553482167e-08, "per_seed": [ 1.9253596406088036e-07, 2.091002073711934e-07, 2.2312131875423802e-07, 1.0322276722263268e-07, 1.9205077705919393e-07, 1.9699885456248012e-07, 1.9496692971188168e-07, 1.638328370745512e-07 ], "n": 8 }, "comparison": { "delta_mean": -1.7760214641171501e-06, "idea_wins": 8, "n_pairs": 8, "per_seed_diffs": [ -1.390778237464474e-06, -1.3703611472237753e-06, -2.6497396135027884e-06, -1.8243816768404033e-06, -2.0408132854754513e-06, -1.9604353553859255e-06, -1.9312202255150623e-06, -1.040442171529321e-06 ], "p_value": 0.0081, "mde": 4.178357032172272e-07, "mde_rel_pct": 21.31270934722776, "verdict": "idea better (significant)", "system_worked": true }, "mechanism_signature": { "trained_model_seed": 0, "baseline_mean_abs_norm_error": 0.0008932742057368159, "idea_mean_abs_norm_error": 1.7583369427143225e-08, "predicted_idea_lower_constraint_error": true, "confirmed": true }, "custom_track": { "name": "sphere_one_step_dynamics", "file": "sphere_track.py", "domain": "geometric_dynamics" } }, "system_verdict": "worked", "practical_verdict": "helps", "mechanism_ok": 1, "system_judged": true }